With AI increasingly making its way into valuation professionals’ workflows, debate has naturally arisen over where to impose boundaries on its role, including in the context of writing valuation reports.
Nancy McCarthy, a professional writer and former editor of The Valuation Examiner, has some valuable thoughts on the use of AI in report writing. “AI is often billed as a writing partner — but it’s not,” she says.
“Good writing is rewriting, calibrating what you say, and that doesn’t change for business writing,” McCarthy says. But that’s not in AI’s wheelhouse.
McCarthy provides editorial consulting and ghostwriting services to business valuators, among others. For the past several years, she’s also presented a segment of Rod Burkert’s well-regarded five-part webinar through NACVA, “Report Writing: Review and Analysis.” She recently shared some of her thoughts on the evolution of AI in valuation report writing.
Recognize Where We Are
According to McCarthy, the valuation industry is moving past the point of arguing over whether valuators should use AI when drafting reports. “The bottom line is that we’re using it.” The critical issue now, she says, is the application of stronger editorial rigor. “AI can be precise in a lot of things, but it still needs oversight.”
McCarthy emphasizes the importance of human supervision and quality control. “In a valuation report, it has to be 100% the numbers telling the story, accurately and clearly. AI isn’t going to do that. It’s going to focus on pulling data and sounding good. You have to guide it and take a critical look at what it spits out.
“As much as AI knows, it’s not a thinking entity. It’s a retrieval entity, and it’s only as good as the information it retrieves. If you let it, it will spew forth whatever’s been programmed into it. It might sound good, but it might not be what you actually want.”
The primary way valuators can guide AI is through their prompts. McCarthy recommends valuators use a prompt Burkert devised:
“Base your answer only on the uploaded documents. If the information isn’t found, say ‘not found in documents.’ For each claim, include the supporting document, the page number, and the exact quote.”
Making AI show its work in this way simplifies your verification process, a mandatory ingredient when using AI tools.
Avoid the Template Trap
Valuators may believe they can sidestep some of the need for ongoing guidance and oversight by uploading report templates to their AI tools. After all, if you show the AI the right format and structure, what could go wrong?
“Templates are wonderful until they’re not,” McCarthy says. Clean templates paired with AI can create exponential efficiency, but poor templates combined with AI can lead to disaster.
“AI will very cheerfully expand whatever’s in a template, including mistakes,” McCarthy says. “It doesn’t find or correct mistakes; it enhances them.”
If, for example, a template contains outdated client data, AI will simply rephrase it and preserve it. “But it’ll be old data that will skew anything you’re writing about,” she says. Similarly, if a template itself is outdated, AI will make it sound better, but the template will still be outdated. Before uploading a template, valuators must ensure it’s clean, with the current information.
“You have to tell AI what to ignore,” McCarthy says. “You also need to tell it each time to update language, check consistency, and rewrite the static sections, such as the mission statement or your biographical information. “You also may need to rebuild charts and graphs and the like to verify assumptions.” Keep in mind, too, that charts and graphs can have meta data from previous engagements.
Set Your Firm Up for Success
As noted, it’s no longer a question of whether valuation professionals will use AI to write their reports. The more relevant question these days is how they use it to write reports. McCarthy encourages firms to be proactive on this matter, starting with governance.
“Firms need to implement frameworks for supervising AI outputs. You can’t just leave it up to chance — you have to have policies and procedures. AI works better when it’s controlled.”
You shouldn’t stop with policies and procedures, though. McCarthy believes AI literacy should be part of the training valuators receive, including what AI can and can’t do for them when writing reports.
“AI can support you with your outlines, summary, and clarity. It can do consistency checks on tone and grammar. It can explain concepts that have neutral descriptions. But it can’t replace your professional judgment or determine formal conclusions.”
In addition, McCarthy wants new AI users to remember that, while good prompts can improve their drafts, weak or outright bad prompts — for example, those that include judgment language, rather than neutral — create risks. And valuators must take the time to document their process, laying out exactly what they used AI for and what they reviewed before relying on AI outputs.
Final Word Ultimately, McCarthy advises valuators to think of AI as an accelerator for reports, not a co-author. “It accelerates valuation writing, but the valuator’s professional judgment remains the engine of the report. AI won’t make weak analysts strong. It will make strong analysts faster.”
